438 Non-invasive sensing of grapevine vegetative status Australian Journal of Grape and Wine Research 22, 438–449, 2016
Calibration of non-invasive fluorescence-based sensors for the manual andon-the-go assessment of grapevine vegetative status in the field
M.P. DIAGO1, C. REY-CARAMES1, M. LEMOIGNE2, E.M. FADAILI2, J. TARDAGUILA1 and Z.G. CEROVIC3,4,5
1 Instituto de Ciencias de la Vid y del Vino, (University of La Rioja, CSIC, Gobierno de La Rioja), 26007 Logroño, Spain; 2 FORCE-A,Université Paris Sud, Orsay, F-91405, France; 3 Laboratoire Écologie Systématique et Évolution, Université Paris-Sud, Unitémixte de Recherche (UMR) 8079, Orsay, F-91405, France; 4 Centre National de la Recherche Scientifique (CNRS), Orsay, F-
91405, France; 5 AgroParisTech, Paris, F-75231, FranceCorresponding author: Dr Maria P. Diago, email [email protected]
Abstract
Background and Aims: Optical sensors can accomplish frequ ent and spatially widespread non-destructive monitoring of plantnutrient status. The main goal was to calibrate a fluorescence sensor, used both manually (MXH) and on-the-go (MXM), for theassessment of the spatial variability in the vineyard of the concentration of chlorophyll, flavonol and nitrogen in grapevine leaves,against that of a leaf-clip type optical sensor (DX4).Methods and Results: Measurements were taken in a commercial vineyard on the adaxial and abaxial sides of leaves of nineVitis vinifera L. cultivars, manually with the DX4 andMXH, and with theMXMmounted on an all-terrain vehicle. A significant cor-relationwas obtained for the chlorophyll and nitrogen indices of MXH and DX4 (R2>0.90) and of MXM and DX4 (R2> 0.74), andthe calibration equations were defined. A similar spatial distribution was achieved for the chlorophyll, flavonol and nitrogen indi-ces of the leaves.Conclusions: The capability of the fluorescence sensor, used manually and on-the-go, for characterising the nutritional status ofgrapevines was demonstrated.Significance of the Study: This work reports the first calibration of the hand-held and on-the-go fluorescence sensor to assesskey nutritional parameters of grapevines. The applicability of this sensor on-the-go to characterise the spatial variability of thevegetative status of a vineyard for the delineation of homogeneous management zones was proved.
IntroductionVineyards have been demonstrated to be spatially variable.Within a vineyard, changes in soil type or depth, slope andexposure may occur. Each individual factor and the interactionamong them influence grapevine development, leading to dif-ferences in vine growth, yield or grape composition (vanLeeuwen 2010). The knowledge and study of the spatial vari-ability of the several features of a vineyard allow a differenti-ated, optimised management, which is known as precisionviticulture. For this purpose, the collection and use of largeamounts of data related to the plant’s physiological status, yieldand grape composition are needed (Proffitt 2006).
Chlorophyll (Chl), flavonols (Flav) and nitrogen (N) arekey physiological constituents of grapevines. Chlorophyll isthe pigment responsible for photosynthesis and increases untilgrapevine leaves are fully expanded and starts to decrease af-terwards, as soon as it attains its maximal value (Kriedemannet al. 1970). Flavonols comprise a class within the flavonoids,a secondary metabolite group of compounds sharing a three-ring phenolic structure. Flavonols in plants display awide rangeof physiological functions, involving microbial interactions(Koes et al. 1994) and free radical scavenging (Markhamet al. 1998), but their most prevalent role appears to be as UVscreening agents (Flint et al. 1985, Smith and Markham1998). Flavonol biosynthesis is upregulated not only becauseof UV radiation but also in response to other biotic and abioticstresses, such as N/phosphorus depletion (Lillo et al. 2008),
1111/ajgw.12228Australian Society of Viticulture and Oenology Inc.
low temperature (Olsen et al. 2009) and salinity/drought stress(Tattini et al. 2004, Agati et al. 2011a). Leaf Chl and Flav con-centration on a surface basis depends on leaf age and theamount of light radiation received during their development.Both increase with leaf expansion and light exposure untilveraison, while afterwards, leaf Chl usually decreases (Louiset al. 2009) while Flav remain unvaryingly high (Downeyet al. 2003). Nitrogen is considered to be one of themost impor-tant factors for biomass production (Lemaire et al. 2008, Agatiet al. 2013a) and grapevine metabolism, as it is crucial for vinedevelopment and fruit yield (Guilpart et al. 2014). Therefore,the assessment of the vineyard Chl and N status is necessaryand helpful to delineate strategies for fertilisation and canopymanagement intended to improve the grapevines’ balanceand fruit composition. In grapevines, excessive N can some-times be even more damaging than N deficiency because vineswould be more prone to disease and insect infestations (Dordas2009). Overfertilisation usually produces grapes of poorer com-position (Keller 2010), and plants are more susceptible to abor-tion of flowers and reduced fruitset (Vasconcelos et al. 2009).
Cartelat et al. (2005) have shown that both flavonol andChl concentration is important for the assessment of the N sta-tus of the plant. This ratio is known as the N balance index(NBI=Chl/Flav), and its relationship with the N status has alsobeen reported by several authors for other species (Tremblayet al. 2012). Chlorophyll concentration increases, whereas thatof flavonol decreases with increased N application, so that the
Diago et al. Non-invasive sensing of grapevine vegetative status 439
NBI increases with N fertilisation. Therefore, in the frameworkof precision farming, the epidermal concentration of flavonolsand leaf Chl is useful for N management (Tremblay et al.2012) as it allows the NBI to be calculated.
Leaf Chl, flavonol and N status are usually analysed withdestructive wet chemistry methods. Compared with the latter,optical methods provide much faster assessment and are theonly ones allowing practical whole plot analysis. Opticalmethods are based on leaf transmittance, reflectance orfluores-cence, and they can be used as proximal sensors. Proximalsensing, which includes all detecting technologies that gatherinformation from an object when the distance between thesensor and the object is less than, or comparable with, someof the dimensions of the sensor, have emerged as an alternativeto remote sensing in viticulture. Proximal sensing provides asuccessful solution to most of the drawbacks, such as large pro-portion of background noise in the images, limited temporalflexibility and elevated cost of aerial monitoring, of remotesensing in vertically trellised vineyards worldwide. Amongthe wide variety of technologies used in proximal sensing, Chlfluorescence has been introduced in viticulture for the moni-toring of anthocyanin accumulation, the assessment of vinevigour and the control of diseases in plants (Agati et al. 2008,Bellow et al. 2012, Latouche et al. 2013). It is possible to obtainestimates of anthocyanin and Flav by using the non-destructiveChl fluorescence excitation screening method: the higher theanthocyanin or Flav concentration in the berry skin or leaf,the lower the Chl fluorescence signal.
Proximal sensors can be either hand-held or mounted ontoa machine, allowing both the acquisition of data in a non-destructive way (Tisseyre 2013). The spatial resolution of thedata recorded, however—that is the number of measurementsper unit plot surface—differs between the manual and the on-the-go operation. While the hand-held sensors are carried byan operator, who takes the measurements, the on-the-go de-vices are mounted onto a motorised vehicle [i.e. tractor andall-terrain vehicle (ATV)] andmeasure automatically accordingto a triggering protocol design. Therefore, the spatial resolutionincreases with the on-the-go sensors, and a large amount ofdata can be recorded in less time (Tisseyre 2013). Furthermore,when a global positioning system is used, the data points ob-tained can be georeferenced and interpolated to generate acomprehensive map of the crop condition (Tremblay et al.2012). Tractor-based mapping would be extremely valuable,as tractors frequently move along the rows to undertake manyvineyard operations. Therefore, by mounting sensors on trac-tors, information could be gathered with no time-cost and atdifferent stages of vine development (Taylor et al. 2005).
Themain goal of the present studywas to calibrate against aleaf-clip optical sensor, used as calibrated reference, and toevaluate the performance of a portable non-destructive fluo-rescence sensor used both manually and on-the-go (on amotorised platform) for the assessment of the spatial variabilityand mapping of the concentration of Chl, Flav and N in grape-vine leaves within a vineyard.
Materials and methods
Site descriptionThe study was undertaken in 2012 during the last week ofSeptember and first week of October at a 1.43ha commercialvineyard owned by the nursery Vitis Navarra located in Vergalijo(Latitude 42°27′45.96″, Longitude 1°48′13.42″, Altitude 325m),Navarra, Spain. The vineyard was planted with nine red
international cultivars: Cabernet Sauvignon, Carmenere,Caladoc, Grenache, Marselan, Maturana Tinta, Pinot Noir,Tempranillo and Syrah. Grapevines were trained to a verticallyshoot-positioned trellis system, with north–south row orienta-tion at 2×1m inter-row and intra-row distances. Grapevineswere planted on Richter 110, with the exception of Tempranillovines,whichwere planted on rootstock 3309. Irrigationwas rou-tinely and uniformly applied across the season for all cultivars.The choice of a vineyard with several genotypes wasmade to in-crease the variability for the development of the calibrationmodels and hence to develop a more robust model.
Fluorescence sensors and indicesThe vineyardwasmonitoredwith three proximal sensors basedon Chl fluorescence: the Multiplex, which was used manually[hand-held Multiplex (MXH)] and on-the-go [Multiplex On-The-Go (MXM)] and the leaf-clip fluorescence sensor, Dualex4(DX4), which served as the reference device.
Leaf-clip fluorescence sensorDualex4 (FORCE-A, Orsay, France), DX4 hereafter, is a leaf-clip sensor with a measuring surface of 6mm diametre, whichmeasures leaf epidermal flavonols by the Chl fluorescencescreeningmethod (Goulas et al. 2004) and the Chl leaf concen-tration by differential transmittance (Cerovic et al. 2012). Itprovides three fluorescence indices: the chlorophyll optical in-dex (CHL) (Equation 1) for the leaf Chl concentration,displayed in Chl units (Cerovic et al. 2012); the flavonol opticalindex (FLAV) (Equation 2.) for the epidermal Flav concentra-tion in absorbance units; and the NBI index (Equation 3), asthe ratio of Chl to Flav (Cartelat et al. 2005), which refers tothe leaf N concentration (Cerovic et al. 2012).
where T850 and T710 are the leaf transmittance at 850and710nm, respectively; FRF is the far-red Chl fluorescence emis-sion (>710nm) excited by red (_R, 650nm) or UV (_UV,375nm) light; and the subscripts AD and AB refer to the adax-ial and abaxial sides of the leaf, respectively.
Instrument calibrationsThe DX4 CHL and FLAV indices have been validated in a previ-ous work by Cerovic et al. (2012) against Chl extracts andDualex3 FLAV index, respectively, and the robustness of thesecalibrations of Dualex4 for the assessment of Chl and Flav ingrapevine leaves is described. In that study, the reproducibilityand accuracy of the calibration were provided, as well as modelstatistics such as residual sum of squares (RSS), root meansquare error (RMSE), bias (BIAS) and standard error of pre-diction corrected (SEPC). The method and technology forthe measurement of Chl and Flav were not changed betweenthe different Dualex versions (DX4, the one used in the pres-ent study, and previous one, Dualex 3). The same light-emit-ting-diode sources and filters (therefore wavelengths) areused. The DX4 NBI index has been previously validated byCartelat et al. (2005) to assess the N status of wheat andrecently by Cerovic et al. (2015) against the N concentrationin grapevine leaves over a period of 5 years. The work of
440 Non-invasive sensing of grapevine vegetative status Australian Journal of Grape and Wine Research 22, 438–449, 2016
Cerovic et al. (2015) demonstrates the robustness of thepredictive capability of the NBI index against the Nconcentration determined by wet chemistry, and values forsensitivity, accuracy and RMSE are provided. In that samework (Cerovic et al. 2015), calibration of DX4 against Chlextracts was also provided and the same model statistics asfor N were shown.
Hand-held fluorescence sensorThe Multiplex (FORCE-A), MXH hereafter, is a hand-held,multi-parametric fluorescence sensor based on light-emitting-diode excitation and filtered-photodiode detection that are de-signed to work in the field under daylight on leaves, fruits andvegetables (Ben Ghozlen et al. 2010). The sensor illuminates asurface of 8 cm diameter at a 10 cm distance from the source.This device provides 12 signals and several signal ratios, amongthem the indices that are the object of the present study: SFR(Equations 4,5), FLAV (Equation 6) and NBI (Equations 7–9),which are defined as
SFR�R ¼ FRF�RRFR
(4)
SFR�G ¼ FRF�GRF�G
(5)
FLAV ¼ logFRF�RFRF�UV
� �(6)
NBI�R ¼ FRF�UVRF�R
(7)
NBI�G ¼ FRF�UVRF�G
(8)
The SFR index is linked to the Chl concentration of leaves.It is a simple fluorescence ratio (SFR) of far-red Chl emission(FRF, 735nm) divided by red Chl emission (FR, 685nm) underred (FRF_R and FR_R, respectively) (Equation 4) or green exci-tation (FRF_G and RF_G, respectively) (Equation 5). Becauseof the overlap of the Chl absorption and emission spec-trum, re-absorption occurs at shorter wavelengths (RF)but not at longer wavelengths (FRF) (Gitelson et al.1999, Pedrós et al. 2010). Therefore, SFR increases withincreasing sample Chl concentration.
The FLAV index (Equation 6) compares the Chl fluo-rescence intensity emitted as far-red fluorescence underultraviolet (FRF_UV) and red excitation (FRF_R), whichrepresents a differential absorption measurement (in ac-cordance with the Beer–Lambert law) that is proportionalto the Flav concentration of the epidermis (Ounis et al.2001, Agati et al. 2011b).
The NBI displayed in Equations 7 and 8 is related to theN status of the plant and proportional to the Chl-to-Flavratio proposed by Cartelat et al. (2005) but simplified. Itutilises only two signals as the ratio of FRF_UV and RF_Rin NBI_R, or green excitation (RF_G) for NBI_G.
Besides the NBI_R and NBI_G given by Equations 7 and 8,respectively, we calculated also the NBI index (NBIC) sepa-rately for the adaxial and the abaxial leaf sides, as well as forthe whole leaf, based on the ratio between the SFR and FLAVindices of the MXH. The NBIT index of Equation 9 is the calcu-lated hand-held Multiplex index that corresponds to the oneobtained with the DX4. It takes into account the total Chl con-centration of the leaf (numerator) and the sum of the epider-mal Flav of the abaxial and adaxial sides of the leaf(denominator).
NBIT ¼ ChlFlav
¼ SFRAD þ SFRAB
FLAVAD þ FLAVAB(9)
Equations 10 and 11 show the formulae for the computa-tion of the NBIC index for the adaxial leaf side.
NBIC RAD ¼ ChlAD
FlavAD¼ SFR_RT
FLAVAD(10)
NBIC GAD ¼ ChlADFlavAD
¼ SFR_GT
FLAVAD(11)
In Equations 12 and 13, the total NBIT index of Equation 9is rewritten explicitly for the red (R) and green (G) excitation inMXH, respectively.
NBIC�RT ¼ ChlFlav
¼ SFR�RAD þ SFR�RAB
FLAVAD þ FLAVAB(12)
NBIC�GT ¼ ChlFlav
¼ SFR�GAD þ SFR�GAB
FLAVAD þ FLAVAB(13)
On-the-go fluorescence sensorThe Multiplex On-The-Go (Multiplex 321 LD, FORCE-A) ormounted Multiplex, hereafter MXM, is a Multiplex sensoradapted to be used mounted on an ATV or a tractor. It issynchronised with a global positioning system that allows forgeoreferencing of the fluorescence measurements. This devicemeasures a surface of 10 cm diameter from a distance of ap-proximately 20cm. The fluorescence signals and indices pro-vided by the MXM are the same as those yielded by the MXH.In this study, these indices included SFR, FLAV and NBI,among others. The leaves measured with the MXM are a mixof AD and AB leaves, even though the prevailing exposed sidewill be the adaxial side of the leaf.
The SFR_R and the FLAV indices are calculated follow-ing Equations 4 and 6, respectively. In addition to NBI_Rand NBI_G given by Equations 7 and 8, respectively,which coincide for both MXH and MXM, the NBI index(NBIC) based on the ratio of SFR and FLAV indices ofthe MXM was also calculated.
An exhaustive description of all formulae and equa-tions of the fluorescence indices provided and calculatedfrom the three sensors, DX4, MXH and MXM can be foundin Table S1.
The comparison among these NBI indices would enableestimation of the error in the NBI provided by the MXM
with respect to the NBI given by the DX4, which is consid-ered the reference. Towards that aim, the indices corre-sponding to the adaxial (named using AD as subscript)and abaxial (named using AB as subscript) sides of the
Diago et al. Non-invasive sensing of grapevine vegetative status 441
leaves, independently, were compared with the total (ad-axial and abaxial) indices (named using T, for total, as sub-script) for the whole leaf.
Fluorescence measurementsThe 24 rows of the vineyard plot under study were manuallymonitored with the MXH, the DX4 and on-the-go by theMXM. For themanual devices, in each row, 13 sampling points,each one comprising three adjacent vines, were defined, at10 m intervals. Measurements with MXH and DX4 wereconducted on the east side of the rows, on 12 leaves persampling point (four leaves per vine). The same leaf wasmeasured once with the MXH and twice with the DX4, onboth sides, abaxial and adaxial. The leaves measured withthe hand-held devices were located at the mid-upperheight of the canopy to satisfy the condition of being atthe same height targeted and measured by the MXM.
A total of 3744 manual measurements (24 rows × 13sampling points × 12 leaves) on the abaxial and 3744 mea-surements on the adaxial sides of leaves with the MXH, and7488 measurement (24 rows × 13 sampling points × 12leaves × two measurements) on each leaf side with theDX4 were taken. All rows were monitored on both sidesof the canopy with the MXM, mounted on an ATV movingat 5 km/h. The MXM was placed 1.5m above the ground,so that the leaves on the mid-upper part of the canopy(those same measured with the manual devices) weremeasured at a 20 cm distance. The acquisition rate forthe MXM was 60Hz.
Data treatment and statistical analysisThe data obtained with the MXH and MXM devices werefiltered by discarding readings higher than 4200mV toavoid possible nonlinearity in the sensor response. OnMXH, the values lower than 10mV, which correspond tothe residual offsets, and the readings with a coefficient ofvariation of the FRF_R signal larger than 20% were alsoremoved because this indicates that the sensor shifted dur-ing measurement acquisition or that fluctuations in vari-able Chl fluorescence were too large. On the MXM, ahistogram was computed to identify the data correspond-ing to leaves or canopy gaps. The latter were removed.
Figure 1. (a) Example of the grid generated to create a common framework for th
After the filtering, the data of the two devices werestandardised against a blue plastic-foil standard (Force-A)in order to compare the data obtained with other sensorsand data collected under other measuring conditions. Priorto any statistical analysis, the data obtained with the threedevices were corrected by applying the standard normalvariate transformation to avoid the influence of measuringon different days (Legendre and Legendre 1998). Outlierswere identified and excluded from the dataset by applyingthe Tukey method (Tukey 1977).
All three devices automatically provide the indices forthe side from which the leaf is measured, adaxial or abax-ial. In the case of DX4 and MXH, to obtain calculated andtotal (T) leaf indices, the indices of the adaxial and theabaxial sides had to be added (cf. Equations 3, 9, 12 and13). Indeed, each side of the leaf, either the palisade (ad-axial) or the spongy mesophyll (abaxial), is different(Vogelmann and Evans 2002) and will have a fluorescenceSFR and FLAV index specific to that side of the leaf. Anexception is the CHLT index of the DX4, which was usedas the average of the adaxial and the abaxial sides (Equa-tion 3) because the DX4 measures the Chl in transmit-tance mode; therefore, regardless the side from whichthe leaf is measured, the Chl index reflects Chl concentra-tion of the whole leaf (Cerovic et al. 2012).
Once data were properly pretreated, the correlationsbetween the same indices obtained by DX4 (referencemethod) and by the MXH were computed and analysed.Correlations were separately calculated for the indices ofthe adaxial and abaxial sides of the leaf and for the totalleaf indices.
The next step involved the calculation of the correlationsbetween the MXM indices and those obtained with the MXH
and DX4. For that purpose, as there were more data from theMXM than from the two hand-held devices, and for differentgeographical coordinates, the data from each devicewere com-bined into a grid, generated by aggregation of the average ofthe nearest points (Figure 1). This grid allowed for having acommon framework to analyse the correlations betweenMXM, MXH and DX4.
Classical global linear correlation models adjusted by ordi-nary least squares (OLS) were computed to analyse the
e Multiplex On-the-Go and (b) the reference, Dualex4 data.
442 Non-invasive sensing of grapevine vegetative status Australian Journal of Grape and Wine Research 22, 438–449, 2016
relationships between the same indices measured with MXM
and with the two hand-held devices, MXH and DX4. Thestrength and direction of the association were indicated by thedetermination coefficient (R2). The RMSE and the coefficientof variation of the RMSE (%RMSE), calculated as the ratio be-tween the RMSE by the mean value, were also computed to il-lustrate the robustness and accuracy of the predictions.Additionally, in order to provide an estimation of the accuracyof the MXH and MXM in predicting Chl, Flav and N concentra-tion in grapevine leaves under field conditions, the total cumu-lative error %RMSEtotal (Equation 14) was calculated, takinginto account the %RMSE for each device against DX4 (Table 1), and the %RMSEvalidation of DX4 against leaf extractsof Chl and Dualex3measurements of Flav from the calibrationsin Cerovic et al. (2012) and against leaf N concentration deter-mined by wet chemistry (Cerovic et al. 2015).
Data pretreatment and statistical analysis were carried outusing the softwares R (R Core Team2012),Microsoft Office Ex-cel 2013 (Microsoft Corporation, Redmond, WA, USA) andStatistica 9 (StatSoft., Tulsa, OK, USA).
Finally, the point-to-point measurements of thefluorescence indices of the three devices were interpolatedto generate a continuous surface. Towards that end, theexperimental variograms were computed and fitted thebest model. The parameters of the fitted variograms, range,sill and nugget, were then used to apply the interpolationmethod of ordinary kriging. The software ArcGis 9.3 (ESRI,Redlands, CA, USA) was used for these geostatisticalanalyses.
Results
Calibration of the hand-held fluorescence sensor (MXH) againstthe leaf-clip sensor (DX4)Chlorophyll indices. The indices related to the Chl concen-tration in leaveswere CHL inDX4 (reference) and SFR inMXH.
Table 1. Calibration linear models for the fluorescence indices derived from the hanthe leaf-clip, Dualex4.
The 95% confidence intervals for the fit coefficients are indicated in brackets.BIAS, bias; CHL, chlorophyll optical index; DX4, Dualex4; FLAV, flavonol opnitrogen balance index; RMSE, root mean square error;%RMSE, coefficient of v(n = 302 for MXH and n = 143 for MXM); RSS, residual sum of squares; SEPC, st
Figure 2 shows the correlations among these indices obtainedfor each side of the leaf (adaxial and abaxial) and for the wholeleaf. Figure 2a evidenced no difference between CHLAD andCHLAB measured by DX4 (n=302, R2=0.99, P< 0.001, nooffset). This was an expected result as DX4 works on transmit-tance mode for Chl; hence, regardless the side of the leaf that ismeasured the whole leaf Chl concentration is determined.From the results in Figure 2b, it can be seen that the indicesSFR_RT and SFR_GT acquired with the MXH were similar withan R2 of 0.99 (n=302, P< 0.001) and that a small bias in favourof G excitation was detected. Figure 2c,d shows a strong posi-tive correlation between the SFR index, obtained from G andR excitation, respectively, with the CHLT index, yielding R2 of0.93 (n=302, P< 0.001) for SFR_GT and R2 of 0.92 (n=302,P< 0.001) for SFR_RT. In both cases, there was an offset of10.4 (Figure 2c) and 15.3 (Figure 2d), which can be traced tothe fluorescence reabsorption method used in the MXH forthe SFR index (FRF/RF). Figure 2e,f illustrates the fact thatthe correlations between the SFR index under red or greenexcitation from the adaxial side with the CHLT index wereequivalent to the correlations of CHLT with the ‘total’SFR_RT (Figure 2d) or SFR_GT (Figure 2c). These resultsindicate that the SFR indices from the adaxial side, whichare reflecting the Chl concentration of the palisade meso-phyll, are the component of the ‘total’ SFRT index thatmostly determined its variability among leaves; therefore,the SFRAD can be used as a proxy of the leaf Chl concen-tration. Furthermore, when comparing the adaxial sidewith the abaxial side of the same SFR index (SFR_G orSFR_R), strong correlations were obtained with R2 of0.77 (n= 302, P< 0.001) (Figure 2g) and of 0.74 (n = 302,P< 0.001) (Figure 2h), respectively. A correlation matrixshowing all possible relationships between the Chl relatedindices of DX4 and MXH is included in Figure S1.
Flavonol indices. The fluorescence index FLAV is related tothe leaf epidermal flavonols. Figure 3 shows the correlationsbetween the FLAV indices measured with the two hand-helddevices, DX4 and MXH. With the DX4 measurements, FLAVAB
d-heldMultiplex andMultiplex On-the-Go sensors using the indices obtained with
Intercept and slope coefficients were significant at P< 0.001 for all models.tical index; MXH, hand held Multiplex; MXM, Multiplex On-the-Go; NBI,ariation of the RMSE calculated as the ratio of the RMSE to themean value.andard error of prediction corrected for bias; SFR, simple fluorescence ratio.
Figure 2. Correlations between the chlorophyll optical indices (CHL) and (SFR)related to the leaf chlorophyll concentration obtained with the hand-heldMultiplex (MXH) and Dualex4 (DX4) sensors. Coefficients of determination (a)R2 = 0.993; (b) R2 = 0.996; (c) R2 = 0.925; (d) R2 = 0.917; (e) R2 = 0.927; (f)R2 = 0.920; (g) R2 = 0.767; (h) R2 = 0.743 were significant at P< 0.001. AD,measurement taken on the adaxial side of the leaf; AB, measurement taken onthe abaxial side of the leaf; T, global index including measurements on adaxialand abaxial sides of the leaf; _R, fluorescence measurements using a redexcitation source; and _G, fluorescence measurements using a greenexcitation source. Dashed line in (b) represents the 1:1 line. Dotted lines in (c)and (d) represent the regression lines of the same slope without the offset ofthe observed correlations. (n = 302).
Figure 3. Correlations between the flavonols optical indices (FLAV) related tothe leaf flavonols concentration obtained with the hand-held Multiplex (MXH)(○) and Dualex4 (DX4) (•) devices. Coefficients of determination (a)R2 = 0.929 (○), R2 = 0.964, (•); (b) R2 = 0.270 (○), R2 0.220 (•); (c)R2 = 0.121 (•), R2 = 0.047 (○); (d) R2 = 0.815 (•); (e) R2 = 0.779 (•) weresignificant at P< 0.001. AD, measurement taken on the adaxial side of the leaf;AB, measurement taken on the abaxial side of the leaf; and T, global indexincluding measurements on adaxial and abaxial sides of the leaf. Dashed linein (e) represents the 1:1 line. n = 302.
Diago et al. Non-invasive sensing of grapevine vegetative status 443
yielded a strong correlation (R2=0.96 at P< 0.001) withFLAVT, which is the sum of FLAVAD and FLAVAB and repre-sents the total amount of Flav in the leaf (Figure 3a), whileFLAVAD correlated poorly with FLAVT (R2 = 0.27 atP< 0.001) (Figure 3b) and FLAVAB (Figure 3c) owing to the re-duced variation of Flav on this side of the leaf, which is moreexposed to sunlight and tends to have a maximum epidermalFlav concentration. The same behaviour can be seen for theFLAV indices obtained with the MXH (Figure 3a–c), wherethe Flav determined on the abaxial side of the leaf (FLAVAB)appears to be those influencing the variation of the total leafFlav (FLAVT), with R2 = 0.93 (P< 0.0001). This fact is alsoproved by the high correlation (R2 = 0.82 at P< 0.0001)obtained between the Flav of the abaxial side of the leaf,measured by the MXH, and the Flav of the whole leaf mea-sured by the DX4 (Figure 3d). The relationship betweenFLAVT measured with DX4 and MXH (Figure 3e) shows
an offset of 1.4 absorbance units, which can be traced tothe difference in wavelengths, and therefore extinctioncoefficients, for Flav used in DX4 and MXH, which were375 nm for DX4 and 385 nm for MXH in the UV regionand 650 nm for DX4 and 630nm for MXH in the red partof the spectrum. A correlation matrix showing all possiblerelationships between the FLAV indices of DX4 and MXH
is included in Figure S2.
Nitrogen indices. Two different NBI indices have beencompared, the NBI index and the NBIC index. The first isthe NBI index provided directly by the MXH, which is com-puted using Equations 7–9. The second (NBIC) is an NBIindex calculated for each side of the leaf or for the wholeleaf (total NBI), using Equations 10–13, which are basedon the formula given by Cartelat et al. (2005). The NBICindex from the MXH (Equation 9) is calculated followingthe same rationale as the NBI (DX4) in Equation 3, providedby the DX4, that is, the ratio of Chl-to-Flav.
Figure 4 shows the correlations among the indices relatedto the N status from the abaxial or the adaxial sides of the leafas well as from the whole leaf, measured by the same deviceor between the two devices. The two ways of calculating theNBI index are taken into account, NBI_R and NBIC_R indices,which are derived when red excitation was used. It should benoted that the global NBIT index from the DX4 strongly corre-lated with either NBIAD (R2=0.98, P<0.001) or NBIAB
Figure 4. Correlations between the nitrogen balance (NBI) indices related to theleaf nitrogen balance obtained with the hand-held Multiplex (MXH) and Dualex4(DX4) devices. Coefficients of determination (a) R2 = 0.978; (b) R2 = 0.966; (c)R2 = 0.903; (d) R2 = 0.874; (e) R2 = 0.918; (f) R2 = 0.926 were significant atP< 0.001. AD, measurement taken on the adaxial side of the leaf; AB,measurement taken on the abaxial side of the leaf; T, global index includingmeasurements on adaxial and abaxial sides of the leaf; and _R, fluorescencemeasurements using a red excitation source. n = 302.
444 Non-invasive sensing of grapevine vegetative status Australian Journal of Grape and Wine Research 22, 438–449, 2016
(R2=0.97, P< 0.001) of the same device (Figure 4a,b) and thatthe determination coefficient between NBIAD and NBIABprovided by the DX4 was also high (R2 = 0.90, P< 0.001)(Figure 4c). Therefore, taking into account that the CHLindex of DX4, either adaxial or abaxial, is measuring the totalChl concentration of the leaf, as stated before, and given thatthe equation to calculate the NBI index of the Dualex for theadaxial and the abaxial sides of the leaf is defined as the ratioof CHL-to-FLAV that corresponds to each side of the leaf (TableS1), the NBI index for adaxial or abaxial sides of the leaf is theratio of the leaf Chl concentration of the whole leaf to the Flavconcentration of the adaxial or abaxial sides of the leaf. Thismeans that the Chl (represented by the CHL index in the nu-merator) is the component that is mostly determining theNBI value, because the flavonol concentration, representedby the FLAV index in the denominator, is the only componentof the ratio that is different. A similar result forMXH can be seenwhen analysing the correlations between the global NBIC_RT
with NBIC_RAB (R2=0.87, P< 0.001, Figure 4d) and NBIC_RAD
(R2=0.92, P< 0.001, Figure 4e). Indices NBIC_RAB andNBIC_RAD have been calculated as in the DX4, the numeratorbeing global SFRT and the denominator being the epidermalflavonols of either the abaxial (FLAVAB) or the adaxial sides(FLAVAD). The strong correlation of the total NBIC_RT withthe indices of the two sides of the leaf, where again the only fac-tor varyingwas the FLAV index, indicated that the Chl concen-tration (SFR index) was the component influencing the NBIindex. The study of the relationships between the abaxial andadaxial versions of the NBI index provided by the MXH
(NBI_RAB and NBI_RAD, respectively) with the abaxial and
adaxial versions of the calculated NBIC index from MXH
(NBIC_RAB and NBIC_RAD) revealed a strong significant corre-lation with R2 higher than 0.88 at P< 0.001 (Figure S3). Whenthe twoNBI indices of theMXH (NBI andNBIC)were comparedwith the reference, a higher value of the determination coeffi-cient corresponded to the NBIC_RT calculated as the Chl index(SFR_RT) divided by the Flav index (FLAVT) with an R2 of 0.93(Figure 4f), while the NBI_RAD and the NBI_RAB yielded deter-mination coefficients of 0.75 and 0.67, respectively. Therefore,in the case of the MXH, the more suitable NBI index for the as-sessment of the N status of the grapevine leaves would be theNBIC_RT.
The NBI_G index was also analysed, and it yielded similarresults (not shown) to the same index under red excitation(NBI_R). The correlationmatrix, however, showing all possiblerelationships between the NBI indices of DX4 and MXH whengreen excitation (_G) was used instead of red excitation (_R),is included in Figure S4.
The comparison of the indices retrieved with the MXH
and DX4 enabled the definition of the calibration equa-tions (Table 1) to transform the SFR index (provided byMXH) into absolute units of Chl and the FLAV index intoabsorbance units and to link the NBIC (computed fromthe MXH measurements) with the NBI index provided byDX4. The power to estimate the CHL, FLAV and NBI indi-ces provided by the DX4 from MXH measurements is givenby the RMSE values in Table 1. Additionally, the accuracywas also estimated from the computation of the %RMSE,which ranged from 3 to 11% (Table 1). The %RMSEtotal
for the prediction of Chl, Flav and N concentration ingrapevine leaves using the MXH, under field conditions,was 18, 15 and 12%, respectively.
Calibration of the on-the-go fluorescence sensor (MXM) versushand-held devicesIn order to ensure the reliability of the MXM to assess theChl, N and Flav concentration in grapevine canopies in adynamic continuous way, calibration between the indicesmeasured with the MXM and hand-held devices is needed.The goal of this calibration is to be able to quantify the lossof explained variance (in terms or R2) when the Multiplexsensor is used mounted on a vehicle instead of manually.With the two hand-held devices (DX4 and MXH), both theabaxial and the adaxial sides of the leaf are accessible formeasurements; therefore, indices for the whole leaf can beobtained. By contrast, the MXM measures the vines from anATV or a tractor in movement at approximately 20cm fromthe canopy leaves. Under these conditions, leaves cannot beselected or manipulated. Therefore, the MXM will target theleaves with the orientation that they show in the canopy atthat precise moment of measurement. The adaxial side ofthe leaves will be exposed to the sensor most of the time,but there will be also some measurements from the exposedabaxial side of leaves.
The relationships between the MXM indices and hand-heldones were studied applying the global regression model OLS.All the correlations were significant at P<0.001 (Table 2).The calibration equations of MXM against DX4 for CHL, FLAVand NBI are listed in Table 1. As with the MXH, the RMSEvalues, illustrating the power to estimate the CHL, FLAV andNBI indices provided by the DX4 from MXM measurements,are shown in Table 1, as well as the %RMSE values, whichranged from 4 to 18% (Table 1). The %RMSEtotal for the pre-diction of Chl, Flav and N concentration in grapevine leaves
Table 2. Global linear models adjusted by ordinary least squares for the chlorophyll optical, flavonol optical and nitrogen balance indices studied derived frommeasurements with Multiplex On-the-Go versus hand-held Multiplex and the leaf-clip Dualex4 sensors.
The root mean square error (RMSE) is shown in brackets for each relationship. All the coefficients of the model are significant at P-value< 0.001 (n = 143).
AD, indices measured only on the adaxial side of the leaf; AB, indices measured only on the abaxial side of the leaves; CHL, chlorophyll optical index; DX4,Dualex4; FLAV, flavonol optical index; MXH, hand held Multiplex; MXM, Multiplex On-the-Go; NBI, nitrogen balance index; NBIC, this index has beencalculated as a division of SFR to FLAV, to differentiate it from the NBI provided by both Multiplex devices (see Table S1 in the Supporting Information);SFR, simple fluorescence ratio; T, indices calculated for the whole leaf, abaxial and adaxial.
Diago et al. Non-invasive sensing of grapevine vegetative status 445
using the on-the-goMXMwas 22, 15 and 19%, respectively. Ingeneral, higher values of R2 for Chl and NBI between theMXM indices with those from MXH and DX4 correspond-ing to the adaxial side in comparison with those fromthe abaxial side were observed, indicating that the MXM
was mainly seeing the adaxial side of the leaves, as ex-pected. The exception was FLAV because the adaxialFLAV had a small span, as mentioned for the correlationsof the hand-held devices (Table 2). The Chl-related SFRindex obtained on-the-go with the MXM proved to be wellcorrelated with global SFRT from MXH (R2 = 0.71 atP< 0.001), and CHL from the reference DX4 (R2 = 0.75 atP< 0.001). Slightly improved correlations were found forthe NBI indices (with R2 between 0.74 and 0.77 atP< 0.001), but more moderate correlations were foundfor FLAV (R2 of 0.56, P< 0.001) (Table 2).
Comparing the relationships between the SFRT indexfrom MXH and the SFR index from MXM with the CHLTfrom the reference DX4, better and more accurate (lowerRMSE and %RMSE, Table 1) correlations were obtainedfor the MXH device (R2 = 0.92, P<0.001, Figure 2d) thanfor MXM (R2 = 0.75, P< 0.001, Table 2). These results indi-cate that a loss of nearly 20% of information occurredwhen the Multiplex operated on-the-go. For the FLAV in-dex, the loss of explained variability when measuring on-the-go instead of manually with the MXH, was about 13%,although the accuracy of the results remained similar, be-tween 3 and 4%, in terms of %RMSE (Table 1). The samecomparison showed a loss of 15% for the NBI_RT or theNBIC_RT indices. For these indices, the accuracy dimin-ished with the MXM (%RMSE =17–18%) (Table 1), butthese can be considered fairly good figures for non-destructive, on-the-go measurements.
Spatial variabilityFigure 5 depicts the krigged maps for the global indices ofthe three sensors. The maps showed a similar spatial dis-tribution for the three indices studied, independently ofthe device used. Different areas could be clearly identified.
In the case of the Chl and N, a higher value can be seen inthe upper-right (northeast) part of the plot, and at the left(west) part of the plot, there can be seen two vertical rowsfollowing a double and a single grapevine row that wereidentified as Tempranillo and Grenache rows; the highestvalues are located at the bottom-right (southeast) part ofthe plot with a narrower part going towards left, this ir-regular shape perpendicular to the grapevine rows direc-tion followed a sharp change in soil characteristics,regardless the grapevine cultivar or clone planted. Flavfollowed an inverse spatial distribution relative to that ofChl and N.
DiscussionOptical sensing systems are suitable tools to providefrequent and spatially widespread monitoring of plant nu-trient status (Muñoz-Huerta et al. 2013). This paper pre-sents the first calibration of the fluorescence-basedMultiplex sensor, used manually (MXH) and on-the-go(MXM), on grapevine leaves against the Dualex4 as thereference. In this work, the MXH and MXM have beenstudied to determine their capability to satisfactorily mea-sure, in the case of the MXH, and to estimate, in the caseof the on-the-go MXM, the grapevine’s leaf Chl, epidermalFlav and N concentration.
The DX4 was chosen as the reference for several reasons.First, its performance to yield a reliable and accurate measure-ment of the Chl concentration (Cerovic et al. 2012), epidermalFlav (Agati et al. 2008, Cerovic et al. 2012) and N concentration(Cerovic et al. 2015) has been shown even in grapevine leaves.Second, the efficiency of leaf extraction by organic solventsmay be a potential problem for the calibration of sensors(Lashbrooke et al. 2010) as well as the operator’s skill, whichis a major source of variability (Cerovic et al. 2012). Casaet al. (2015) reported higher average coefficient of variationvalues for Chl assessment using wet chemistry methods thanusing DX4 in four different crops.
The MXH has been used for the study of leaf Flav (Agatiet al. 2011b,Müller et al. 2013) andN status of different species,
Figure 5. Interpolated surfaces by ordinary kriging of chlorophyll indices (CHL and SFR), epidermal flavonol index (FLAV) and nitrogen balance index (NBI) acquiredby Dualex4, hand-held Multiplex and Multiplex On-the-Go sensors. Maps were represented by quantiles.
446 Non-invasive sensing of grapevine vegetative status Australian Journal of Grape and Wine Research 22, 438–449, 2016
such as potato (Ben Abdallah and Goffart 2012), turf-grasses(Agati et al. 2013b), rice (Li et al. 2013) and maize (Zhangand Tremblay 2010, Zhang et al. 2012, Longchamps andKhosla 2014). While its application on vineyards has beenwidely focused on grape bunches to assess their anthocyaninconcentration (Ben Ghozlen et al. 2010, Baluja et al. 2012,Agati et al. 2013a), few studies have been carried out on grape-vine leaves (Serrano et al. 2011). In other crops, the only workreporting the assessment of Chl by MXH was conducted byTremblay et al. (2012), who validated the SFR_GAD indexagainst Chl extractions for kiwi leaves. In the present study,the two SFR indices (_R or _G) have been calibrated againstthe CHLT index provided by the DX4 and the calibration equa-tions for the global SFR_RT and SFR_GT are obtained. Thismeans that the SFR index provided by the MXH can now betranslated, going through DX4 units, into Chl units for grape-vine leaves.
To be able to calculate the Chl and Flav indices at the wholeleaf level, the two sides of the leaf have been measured. This isimportant as the incident light does not penetrate into the en-tire leaf depth; therefore, the emitted Chl fluorescence wouldcorrespond to either the part of the leaf closer to the adaxial sideor the part of the leaf closer to the abaxial side, that is, the pal-isade or spongymesophyll (Koizumi et al. 1998, Karabourniotiset al. 2000, Vogelmann and Evans 2002). The evidence of cor-relations between SFRAB and SFRAD in Figure 2g,h would en-able the calculation of one of them (i.e. SFRAB) from theother one (SFRAD), and in this way, the calibration and correla-tion with theMXH andMXM data would be simplified by usingonly SFRAD. Indeed, adding the SFRAB data to SFRAD data doesnot improve the correlation with CHLT obtained by DX4, but itdoes not make it worse either (Figure S1). The correlation ofSFRAD (R2=0.920 and 0.927, for R and G, respectively) withCHLT (DX4) is much better than SFRAB (R2=0.726 and0.752, for R and G, respectively). It is especially evident forthe higher leaf Chl concentration (above 25μg/cm2, FigureS1) present on the southeast part of the plot, in which theSFRAD versus SFRAB correlation is also poorer (Figure 2g). Thisis the first publication, however, of MXH calibration for the
assessment of the Chl concentration. It would be better if itcould be more generally applicable to all species and Chl distri-bution: dicots, monocots, homogeneous or heterogeneous Chldistribution. This is why we favour the sum of both sides ofSFR. The major difference of the fluorescence-ratio techniquefor Chl assessment compared with transmission-based tech-niques is that it can reveal the differences in dorso-ventral dis-tribution of leaf Chl.
In addition to the Chl-related SFR index, the FLAV indexhas also proved to be able to assess the epidermal Flav concen-tration of grapevine leaves.
On-the-go monitoring of the relevant parameters of a vine-yard related to the vigour and nutritional status of a plant in anon-destructive, fast, and reliable way would enable the map-ping and characterisation of the spatial-temporal variability ofthese variables. This information can help to: (i) optimise vine-yard management (reduction of inputs and other managementcosts and application of variable fertilisation rates) and makingit more sustainable; (ii) identify homogeneous managementzones within a vineyard; and (iii) improve the quality of grapesand wine. For these reasons, the calibration and evaluation ofperformance of the on-the-go MXM device against a widelyverified reference, such as theDX4, to characterise the Chl, epi-dermal Flav and N concentration in grapevine leaves, wereattempted.
The results obtained in the present work showed that thefluorescence indices measured with the MXM successfully ex-plained a significant fraction of the variance of the same indicesmeasured by the DX4, therefore confirming the capability oftheMultiplex sensor operating on-the-go to assess the Chl, epi-dermal Flav and N concentration in grapevine leaves. Severalfactors can affect the fluorescence measurements made in acontinuous way with the MXM. All involve leaf features, suchas the side of the leaf exposed to the sensor, leaf exposure tosunlight during growth and leaf age (i.e. leaf position on theshoot). These three factors are less controlled in on-the-go op-erations in comparison with manual measurements, as in thelatter, the leaves to be measured are susceptible to be chosenand detected under a perpendicular geometry. These factors
Diago et al. Non-invasive sensing of grapevine vegetative status 447
are additional sources of uncertainty for predicting Chl, Flavand N concentration in grapevine leaves with the MXM
compared with that of the MXH and responsible for thediminishment in the accuracy observed. Nevertheless, valuesof%RMSEtotal around 20% for the estimation of these three leafconstituents with the MXM, which measures non-destructivelyand continuously from a vehicle moving at 5km/h, are satisfac-tory and enable the delineation of zones of homogeneousnutritional status and vigour within a given vineyard. This isthe ultimate goal and most powerful capability of such an on-the-go sensor, as it can efficiently provide helpful and robustenough information to the grapegrower from the acquisition ofa large amount of measurements needed for mapping.
The MXM indices, even though they are mainly from theadaxial side, were highly correlated to those from the refer-ence DX4. Light exposure of the measured leaves affects thefluorescence indices indirectly, by impacting the leaf massper area (LMA). Sun-exposed leaves tend to be thicker thanshaded leaves and possess a higher concentration in surface-based Chl (Lichtenthaler et al. 2007). The NBI index, be-cause it is defined as the Chl-to-Flav ratio – the latter beinga surrogate of the LMA (Meyer et al. 2006) – will not beaffected by the increase in LMA. The third effect that influ-ences the MXM measurements would be the leaf age, asthe sensor is measuring the leaves on both primary and sec-ondary (lateral) shoots. Leaves on primary shoots are olderand thicker than those on secondary shoots. Similarly tothe light exposure effect, the NBI, which is a ratio of Chland Flav, will be less affected by the leaf age than simplesurface-based indices such as SFR and FLAV.
The indices related to the epidermal concentration ofFlav and Chl are valuable for N management (Tremblayet al. 2012) as they allow the calculation of the NBI index.The latter, as an indicator of the N status of the plant, hasbeen analysed here by comparing different sensors andways to calculate it. The NBIC is based on the inverse depen-dence of N on Chl and Flav, which increases the dynamicrange, lessens the leaf position and exposure influenceand, finally, voids the effect of the LMA. In contrast, theNBI index provided by MXH or MXM is not a ratio of theChl to the Flav concentration but the ratio of two signals(FRF and RF), and it can be obtained for either the adaxialor the abaxial sides of the leaf but not for the whole leaf. Be-sides these differences in the equations defining the NBI in-dex, the NBI provided directly by the MXM was equally goodas the calculated NBIC, of the same device compared withthat of the DX4 reference. This finding has important practi-cal implications when using the MXH and MXM devices. Forthe MXH, the NBIC would be a more accurate index of the Nstatus of the plant, while when MXM is employed, bothways of calculating the N balance index can be equally usedto estimate the variation in the N status within the vineyard.
Efficient mapping of leaf attributes was demonstratedin this work. Therefore, in the framework of precisionfarming, the MXM enables a fast, non-destructive, reliableon-the-go assessment of the spatial and temporal variabil-ity (as several measurements may be conducted duringthe season) of important indicators of the vegetative andnutritional status of the grapevine. In this way, detectionof chlorotic vines, susceptible to additional iron or othermineral amendments, may be carried out early in the sea-son, and objective appraisal of the recovery of the plantsafter mineral spraying can be performed with the MXM.
The MXM also allows the assessment of the total Flav,which provides information about the exposure of leaves
to light and their potential susceptibility to diseases orpathogens (Agati et al. 2008, 2013b).
The krigged sufaces have revealed that the indices mea-sured using the Multiplex device, manually and on-the-go,showed the same spatial variability as the indices measuredby the DX4. Both MXH and MXM were also able to bring outthe soil variability within the vineyard plot and its effect onthe vegetative growth of the plant, as well as some variation in-duced by the nature of the planted grapevine cultivar, just asdid the reference DX4. The sharp soil change detected in thesouthmiddle part of the plotwas obvious during fieldmeasure-ments. This change in the soil characteristics was certainlyaffecting the leaf Chl, Flav and N concentration (Van Leeuwen2010); therefore, it was expected to be prompted by the fluo-rescence indices.
Diagnosis of plant N status is valuable for rational manage-ment of N in a sustainable fertilisation context. While theMXM
allows a rapid estimation of the N status distribution within thevineyard and delineation of homogeneous subzones for the ap-plication of variable rate fertilisation strategies, precise quantifi-cation of the N concentration should be conducted either withMXH or DX4 once the homogeneous management zones havebeen defined.
Overall, the Multiplex sensor, used manually or on-the-go,may be used as a reliable phenotyping tool. In the case of theMXM, phenotyping can be carried out in a faster and morecontinuous way, enabling a rapid, reliable and non-destructiveassessment of the spatial variability of the vegetative andnutritional status within a vineyard at several times withinthe season.
ConclusionsOptical sensors are capable of providing numerous andspatially widespread monitorings of plant nutrient statusin comparison with destructive time-consuming wetchemistry analyses. An exhaustive calibration of thefluorescence-based Multiplex optical sensor, used eithermanually or on-the-go, was conducted against the refer-ence Dualex sensor for the first time, to assess the Chland Flav concentration as well as the N status of grapevineleaves.
The capability to satisfactorily measure the nutritionaland vegetative attributes of grapevines using the Multiplexwas demonstrated through the defined calibration equa-tions between the MXH or the MXM and the DX4. Fromthe on-the-go approach, the fluorescence indices mea-sured with the MXM successfully explained a high fractionof the variance of the same indices measured by the DX4,hence confirming the capability of the Multiplex used on-the-go to estimate the Chl, epidermal Flav and N concentration ingrapevine leaves in motion. Of the several, not-controlled fac-tors in on-the-go operations, potentially affecting the perfor-mance of the MXM, the side of the leaf exposed to the sensordid not appear to alter any of the fluorescence indices obtained,while leaf exposure to sunlight during its growth and leaf agewere found to influence more the simple surface-based indicessuch as SFR and FLAV than the ratio-based NBI.
Non-destructive, on-the-go monitoring of key parametersof a vineyard related to the vigour and nutritional status ofthe plant with theMXMwill enable themapping and character-isation of the spatial-temporal variability of these parameters.This information will be valuable to support decision-makingfor optimised vineyard management as well as to delineatehomogeneous zones within the vineyard, in the frame of preci-sion viticulture.
448 Non-invasive sensing of grapevine vegetative status Australian Journal of Grape and Wine Research 22, 438–449, 2016
AcknowledgementsWe would like to thank Vitis Navarra for their help with thefield measurements. Also, we thankMs Borja Millán, Mr Guil-laume Masdoumier and Mr Victor Sicilia for collecting fielddata. This work received funding from the EuropeanCommunity’s Seventh Framework Program (FP7/2007–2013)under Grant Agreement FP7-311775, Project Innovine.
Conflict of interestThe authors declare the following competing financial interest(s):Mr Z. C. Cerovic declares a double link to the FORCE-Acompany: as one of the co-authors of the Dualex patent thatthe company exploits and as a part-time consultant to the com-pany. Other authors declare no competing financial interests.
ReferencesAgati, G., Cerovic, Z.G., Marta, A.D., Di Stefano, V., Pinelli, P., Traversi,M.L. and Orlandini, S. (2008) Optically-assessed preformed flavo-noids and susceptibility of grapevine to Plasmopara viticola underdifferent light regimes. Functional Plant Biology 35, 77–84.
Agati, G., Cerovic, Z.G., Pinelli, P. and Tattini, M. (2011a) Light-inducedaccumulation of ortho-dihydroxylated flavonoids as non-destructively monitored by chlorophyll fluorescence excitation tech-niques. Environmental and Experimental Botany 73, 3–9.
Agati, G., Bricolti, S., Guidi, L., Ferrini, F., Fini, A. and Tattini, M.(2011b) The biosynthesis of flavonoids is enhanced similarly by UVradiation and root zone salinity in L. vulgare leaves. Journal of PlantPhysiology 168, 204–212.
Agati, G., Foschi, L., Grossi, N., Guglielminetti, L., Cerovic, Z.G. andVolterrani, M. (2013a) Fluorescence-based versus reflectance proxi-mal sensing of nitrogen content in Paspalum vaginatum and Zoysiamatrella turfgrasses. European Journal of Agronomy 45, 39–51.
Agati, G., D’Onofrio, C., Ducci, E., Cuzzola, A., Remorini, D., Tuccio, L.,Lazzini, F. and Mattii, G. (2013b) Potential of a multiparametric opti-cal sensor for determining in situ thematurity components of red andwhite Vitis vinifera wine grapes. Journal of Agricultural and FoodChemistry 61, 12211–12218.
Baluja, J., Diago, M.P., Goovaerts, P. and Tardaguila, J. (2012) Spatio-temporal dynamics of grape anthocyanin accumulation in aTempranillo vineyard monitored by proximal sensing. AustralianJournal of Grape and Wine Research 18, 173–182.
Bellow, S., Latouche, G., Brown, S.C., Poutaraud, A. and Cerovic, Z.G.(2012) In vivo localization at the cellular level of stilbene fluorescenceinduced by Plasmopara viticola in grapevine leaves. Journal of Experi-mental Botany 63, 3697–3708.
Ben Abdallah, F. and Goffart, J. (2012) Potential indicators based onleaf flavonoids content for the evaluation of potato crop nitrogen sta-tus. Stafford, J.V., ed. Proceedings of the 11th international confer-ence on precision agriculture; 15–18 July 2012; Indianapolis, IN,USA (The International Society of Precision Agriculture: Monticello,IL, USA) pp. 1–18.
Ben Ghozlen, N., Cerovic, Z.G., Germain, C., Toutain, S. and Latouche,G. (2010) Non-destructive optical monitoring of grape maturation byproximal sensing. Sensors 10, 10040–10068.
Cartelat, A., Cerovic, Z.G., Goulas, Y., Meyer, S., Lelarge, C., Prioul, J.L.,Barbottin, A., Jeuffroy, M.H., Gate, P., Agati, G. and Moya, I. (2005)Optically assessed contents of leaf polyphenolics and chlorophyll asindicators of nitrogen deficiency in wheat (Triticum aestivum L.) FieldCrops Research 91, 35–49.
Casa, R., Castaldi, F., Pascucci, S. and Pignatti, S. (2015) Chlorophyll es-timation in field crops: an assessment of handheld leaf meters andspectral reflectance measurements. Journal of Agricultural Science153, 876–890.
Cerovic, Z.G., Masdoumier, G., Ghozlen, N.B. and Latouche, G. (2012)Anewoptical leaf-clipmeter for simultaneous non-destructive assess-ment of leaf chlorophyll and epidermal flavonoids. PhysiologiaPlantarum 146, 251–260.
Cerovic, Z.G., Ben Ghozlen, N., Milhade, C., Obert, M., Debuisson, S.and Le Moigne, M. (2015) Non-destructive diagnostic test for nitro-gen nutrition of grapevine (Vitis vinifera L.) based on Dualex leaf-clipmeasurements in the field. Journal of Agricultural and Food Chemis-try 63, 3669–3680.
Dordas, C. (2009) Role of nutrients in controlling plant diseases in sus-tainable agriculture: a review. Lichtfouse, E., Navarrete,M., Debaeke,
P., Véronique, S. and Alberola, C., eds. Sustainable agriculture(Springer: Dordrecht, The Netherlands) pp. 443–460.
Downey,M.O., Harvey, J.S. and Robinson, S.P. (2003) Synthesis of fla-vonols and expression of flavonol synthase genes in developing grapeberries of Shiraz and Chardonnay (Vitis vinifera L.) Australian Journalof Grape and Wine Research 9, 110–121.
Gitelson, A.A., Buschmann, C. and Lichtenthaler, H.K. (1999) The chloro-phyllfluorescence ratio F735/F700 as an accuratemeasure of the chloro-phyll content in plants. Remote Sensing of Environment 69, 296–302.
Goulas, Y., Cerovic, Z.G., Cartelat, A. and Moya, I. (2004) Dualex: anew instrument for field measurements of epidermal ultraviolet ab-sorbance by chlorophyll fluorescence. Applied Optics 43, 4488–4496.
Guilpart, N., Metay, A. and Gary, C. (2014) Grapevine bud fertility andnumber of berries per bunch are determined by water and nitrogenstress around flowering in the previous year. European Journal ofAgronomy 54, 9–20.
Karabourniotis, G., Bornman, J.F. and Nikolopoulos, D. (2000) A possi-ble optical role of the bundle sheath extensions of the heterobaricleaves of Vitis vinifera and Quercus coccifera. Plant, Cell & Environment23, 423–430.
Keller, M. (2010) Managing grapevines to optimise fruit developmentin a challenging environment: a climate change primer for viticultur-ists. Australian Journal of Grape and Wine Research 16, 56–69.
Koes, R.E., Quattrocchio, F. and Mol, J.N.M. (1994) The flavonoid bio-synthetic pathway in plants: function and evolution. BioEssays 16,123–132.
Koizumi, M., Takahashi, K., Mineuchi, K., Nakamura, T. and Kano, H.(1998) Light gradients and the transverse distribution of chlorophyllfluorescence in mangrove and camellia leaves. Annals of Botany 81,527–533.
Kriedemann, P., Kliewer,W.M. and Harris, J. (1970) Leaf age and pho-tosynthesis in Vitis vinifera L. Vitis 9, 97–104.
Lashbrooke, J.G., Young, P.R., Strever, A.E., Stander, C. and Vivier, M.A. (2010) The development of amethod for the extraction of caroten-oids and chlorophylls fromgrapevine leaves and berries forHPLCpro-filing. Australian Journal of Grape and Wine Research 16, 349–360.
Latouche, G., Bellow, S., Poutaraud, A., Meyer, S. and Cerovic, Z.G.(2013) Influence of constitutive phenolic compounds on the responseof grapevine (Vitis vinifera L.) leaves to infection by Plasmopara viticola.Planta 237, 351–361.
Legendre, P. and Legendre, L. (1998) Numerical ecology (Elsevier:Amsterdam, The Netherlands).
Lemaire, G., Jeuffroy, M.H. and Gastal, F. (2008) Diagnosis tool forplant and crop N status in vegetative stage: theory and practices forcrop N management. European Journal of Agronomy 28, 614–624.
Li, J.W., Zhang, J.X., Zhao, Z., Lei, X.D., Xu, X.L., Lu, X.X., Weng, D.L.,Gao, Y. and Cao, L.K. (2013) Use of fluorescence-based sensors to de-termine the nitrogen status of paddy rice. The Journal of AgriculturalScience 151, 862–871.
Lichtenthaler, H.K., Ač, A., Marek, M.V., Kalina, J. and Urban, O.(2007) Differences in pigment composition, photosynthetic ratesand chlorophyll fluorescence images of sun and shade leaves of fourtree species. Plant Physiology and Biochemistry 45, 577–588.
Lillo, C., Lea, U.S. and Ruoff, P. (2008) Nutrient depletion as a key fac-tor formanipulating gene expression and product formation in differ-ent branches of the flavonoid pathway. Plant, Cell and Environment31, 587–601.
Longchamps, L. and Khosla, R. (2014) Early detection of nitrogen vari-ability in maize using fluorescence. Agronomy Journal 106, 511–518.
Louis, J., Meyer, S., Maunoury-Danger, F., Fresneau, C., Meudec, E.and Cerovic, Z.G. (2009) Seasonal changes in optically assessed epi-dermal phenolic compounds and chlorophyll contents in leaves ofsessile oak (Quercus petraea Matt.(Liebl.)): towards signatures of phe-nological stage. Functional Plant Biology 36, 732–741.
Markham, K.R., Tanner, G.J., Caasi-Lit, M., Whitecross, M.I., Nayudu,M. and Mitchell, K.A. (1998) Possible protective role for 3′,4′-dihydroxyflavones induced by enhanced UV-B in a UV tolerant ricecultivar. Phytochemistry 49, 1913–1919.
Meyer, S., Cerovic, Z.G., Goulas, Y., Montpied, P., Demotes-Mainard,S., Bidel, L.P.R., Moya, I. and Dreyer, E. (2006) Relationships be-tween optically assessed polyphenols and chlorophyll contents, andleaf mass per area ratio in woody plants: a signature of the carbon–nitrogen balance within leaves? Plant, Cell & Environment 29,1338–1348.
Diago et al. Non-invasive sensing of grapevine vegetative status 449
Müller, V., Lankes, C., Schmitz-Eiberger, M., Noga, G. andHunsche, M. (2013) Estimation of flavonoid and centellosideaccumulation in leaves of Centella asiatica L. Urban bymultiparametric fluorescence measurements. Environmentaland Experimental Botany 93, 27–34.
Muñoz-Huerta, R., Guevara-Gonzalez, R., Contreras-Medina, L.,Torres-Pacheco, I., Prado-Olivarez, J. and Ocampo-Velazquez,R. (2013) A review of methods for sensing the nitrogen statusin plants: advantages, disadvantages and recent advances.Sensors 13, 10823–10843.
Olsen, K.M., Slimestad, R., Lea, U.S., Brede, C., Løvdal, T., Ruoff,P., Verheul, M. and Lillo, C. (2009) Temperature and nitrogeneffects on regulators and products of the flavonoid pathway: Ex-perimental and kinetic model studies. Plant, Cell and Environ-ment 32, 286–299.
Ounis, A., Cerovic, Z.G., Briantais, J.M. and Moya, I. (2001) Dual-excitation FLIDAR for the estimation of epidermal UV absorp-tion in leaves and canopies. Remote Sensing of Environment76, 33–48.
Pedrós, R., Goulas, Y., Jacquemoud, S., Louis, J. and Moya, I.(2010) FluorMODleaf: a new leaf fluorescence emission modelbased on the PROSPECT model. Remote Sensing of Environ-ment 114, 155–167.
Proffitt, T. (2006) Precision viticulture: a new era in vineyardmanagement and wine production (Winetitles: Ashford, SA,Australia).
R Core Team (2012) R: A language and environment for statisticalcomputing (R Foundation for Statistical Computing: Vienna,Austria).
Serrano, E., Dias, F., Biais, T. and Dufourcq, T. (2011) Les nouvellestechnologies pour renseigner du statut azote des raisins. Recherchede modeles de prediction a l’aide du capteur multiplex. Actes ducolloque Mondiaviti, Vinitech, 1–2 December 2010; Bordeaux,France (Institut Français de la Vigne et du Vin: Le Grau du Roi,France) pp. 24.
Smith, G.J. andMarkham, K.R. (1998) Tautomerism of flavonol gluco-sides: relevance to plant UV protection and flower colour. Journal ofPhotochemistry and Photobiology A: Chemistry 118, 99–105.
Tattini, M., Galardi, C., Pinelli, P., Massai, R., Remorini, D. and Agati, G.(2004)Differential accumulation offlavonoids andhydroxycinnamatesin leaves of Ligustrum vulgare under excess light and drought stress.New Phytologist 163, 547–561.
Taylor, J., Tisseyre, B. and Praat, J.-P. (2005) Bottling good information:mixing tradition and technology in vineyards. Proceedings of the 7thfruit nut and vegetable production engineering symposium. FRUTIC05 conference. Information and technology for sustainable fruit andvegetable production; 12–16 September 2005;. Montpellier, Francepp. 719–735.
Tisseyre, B. (2013) Sensing systems.embedded in machines: towards abetter management of operations in viticulture. Poni, S., ed.. ActaHorticulturae 978, 21–36.
Tremblay, N., Wang, Z. and Cerovic, Z. (2012) Sensing crop nitrogenstatus with fluorescence indicators. a review. Agronomy for Sustain-able Development 32, 451–464.
Tukey, J.W. (1977) Exploratory data analysis (Addison-Wesely:Boston, USA).
Van Leeuwen, C. (2010) Terroir: the effect of the physical envi-ronment on the vine growth, grape ripening and wine sensoryattributes. Reynolds, A.G., ed. Managing wine quality. viticul-ture and wine quality (Woodhead Publishing: Cambridge,England) pp. 273–315.
Vasconcelos, M.C., Greven, M., Winefield, C.S., Trought, M.C.T. andRaw, V. (2009) The flowering process of Vitis vinifera: a review.American Journal of Enology and Viticulture 60, 411–434.
Vogelmann, T.C. and Evans, J.R. (2002) Profiles of light absorption andchlorophyll within spinach leaves from chlorophyll fluorescence.Plant, Cell & Environment 25, 1313–1323.
Zhang, Y. and Tremblay, N. (2010) Evaluation of the Multiplexfluorescence sensor for the assessment of corn nitrogen status.Khosla, R., ed. Proceedings of the 10th International conferenceon precision agriculture, 18–21 July 2010; Denver CO, USA(International Society of Precision Agriculture: Monticello, IL,USA) p. 9.
Zhang, Y., Tremblay, N. and Zhu, J. (2012) A first comparison ofMulti-plex® for the assessment of corn nitrogen status. Journal of Food,Agriculture & Environment 10, 1008–1016.
Manuscript received: 19May 2015
Revised manuscript received: 9 December 2015
Accepted: 15December 2015
Supporting informationAdditional Supporting Informationmay be found in the onlineversion of this article at the publisher’s web-site: http://onlinelibrary.wiley.com/doi/10.1111/ajgw.12228/abstract
Table S1. Nomenclature and equations of the different chloro-phyll, flavonol and nitrogen balance indices provided by theDualex4 (DX4), hand-held Multiplex (MXH) and MultiplexOn-the-Go (MXM) sensors and calculated from the providedindices.Figure S1. Correlation matrix between the chlorophyll indicesof the hand-held Multiplex (MXH) and the leaf-clip Dualex4(DX4). CHLAD and CHLAB are the chlorophyll indices for adax-ial and abaxial sides of the leaf, respectively, given by the DX4device, and CHLT is the sum of the adaxial and abaxial sides.The simple fluorescence ratio (SFR_G and SFR_R) is the chlo-rophyll index yielded by the MXH. Adaxial (SFR_RAD andSFR_GAD) and abaxial (SFR_RAB and SFR_GAB) sides are alsotaken into account and the index of the whole leaf as the sumof each side of the leaf. Each combination R2 is indicated inthe right side of the graphic diagonal, and all of them are statis-tically significant at P < 0.001 (n = 302).Figure S2. Correlation matrix between the epidermal flavonolindices of the hand-held multiplex (MXH) and the leaf-clipDualex4 (DX4). FLAVAD and FLAVAB are the epidermal flavo-nol indices for adaxial and abaxial sides of the leaf, respectively,and FLAVT is the sum of the adaxial and abaxial sides. Eachcombination R2 is indicated in the right side of the graphicdiagonal, and all of them are statistically significant at P < 0.001(n = 302).Figure S3. Correlation matrix between the nitrogen balanceindices of the hand-held multiplex (MXH) and the leaf-clipDualex4 (DX4). NBI_RAD and NBI_RAB are the nitrogenbalance indices for adaxial and abaxial sides of the leaf,respectively. NBIC_RAD and NBIC_RAB are the nitrogenbalance indices calculated as SFR_R divided to FLAVAD orFLAVAB, respectively. Each combination R2 is indicated inthe right side of the graphic diagonal, and all of them arestatistically significant at P < 0.001 (n = 302).Figure S4. Correlation matrix between the nitrogen bal-ance indices of the hand-held multiplex (MXH) and theleaf-clip Dualex4 (DX4). NBI_GAD and NBI_GAB are the ni-trogen balance indices for adaxial and abaxial sides of theleaf, respectively. NBIC_GAD and NBIC_GAB are the nitro-gen balance indices calculated as SFR_G divided to FLAVAD
or FLAVAB, respectively. Each combination R2 is indicatedin the right side of the graphic diagonal, and all of themare statistically significant at P < 0.001 (n = 302).
Table S1. Nomenclature and equations of the different chlorophyll, flavonol and nitrogen balance indices provided by the three sensors: DX4, MXH and MXM,
and calculated from the provided indices.
CHL DX4 Adaxial (AD) Abaxial (AB) TOTAL
Provided 𝐶𝐻𝐿 𝐶𝐻𝐿
Calculated 𝐶𝐻𝐿 = 𝐶𝐻𝐿 + 𝐶𝐻𝐿
2
MXH Red excitation (R) Green excitation (G) Adaxial (AD) Abaxial (AB) TOTAL Adaxial (AD) Abaxial (AB) TOTAL
Provided 𝑆𝐹𝑅_𝑅 = 𝐹𝑅𝐹_𝑅𝑅𝐹_𝑅
𝑆𝐹𝑅_𝑅 = 𝐹𝑅𝐹_𝑅𝑅𝐹_𝑅
𝑆𝐹𝑅_𝐺 = 𝐹𝑅𝐹_𝐺𝑅𝐹_𝐺
𝑆𝐹𝑅_𝐺 = 𝐹𝑅𝐹_𝐺𝑅𝐹_𝐺
Calculated 𝑆𝐹𝑅_𝑅 = 𝑆𝐹𝑅_𝑅 + 𝑆𝐹𝑅_𝑅
𝑆𝐹𝑅_𝐺 = 𝑆𝐹𝑅_𝐺 + 𝑆𝐹𝑅_𝐺
MXM Red excitation (R) Green excitation (G)
Provided 𝑆𝐹𝑅_𝑅 = 𝐹𝑅𝐹_𝑅𝑅𝐹_𝑅
𝑆𝐹𝑅_𝐺 = 𝐹𝑅𝐹_𝐺𝑅𝐹_𝐺
FLAV DX4 Adaxial (AD) Abaxial (AB) TOTAL
Provided 𝐹𝐿𝐴𝑉 𝐹𝐿𝐴𝑉
Calculated 𝐹𝐿𝐴𝑉 = 𝐹𝐿𝐴𝑉 + 𝐹𝐿𝐴𝑉
MXH Adaxial (AD) Abaxial (AB) TOTAL
Provided 𝐹𝐿𝐴𝑉 = log𝐹𝑅𝐹_𝑅
𝐹𝑅𝐹_𝑈𝑉 𝐹𝐿𝐴𝑉 = log
𝐹𝑅𝐹_𝑅𝐹𝑅𝐹_𝑈𝑉
Calculated 𝐹𝐿𝐴𝑉 = 𝐹𝐿𝐴𝑉 + 𝐹𝐿𝐴𝑉
MXM
(provided) 𝐹𝐿𝐴𝑉 = log
𝐹𝑅𝐹_𝑅𝐹𝑅𝐹_𝑈𝑉
NBI DX4 Adaxial (AD) Abaxial (AB) TOTAL
Provided 𝑁𝐵𝐼 =
𝐶𝐻𝐿𝐹𝐿𝐴𝑉
𝑁𝐵𝐼 = 𝐶𝐻𝐿𝐹𝐿𝐴𝑉
Calculated 𝑁𝐵𝐼 = 𝐶𝐻𝐿 + 𝐶𝐻𝐿 2⁄𝐹𝐿𝐴𝑉 + 𝐹𝐿𝐴𝑉
MXH Red excitation (R) Green excitation (G)
Adaxial (AD) Abaxial (AB) TOTAL Adaxial (AD) Abaxial (AB) TOTAL
